Server-Side goes Serverless

Slides from my talk at Google I/O 2017 Extended Fortaleza where we try to understand why the serverless architecture's been drawing so much attention from developers lately by seeing how this model can help significantly reduce the operational costs and complexity of our applications.

Transcript

1. server-side goes serverless @ythecombinator @feliborgez
2. !!!
3. #1 it has nothing to do with serveless but…
4. #1
5. #2 + architecture - code
6. #3 pretty much focused on experiences
7. stalkr
9. Travis CI 
10. GitHub 
11. Apex 
12. Heroku 
14. http://getstal.kr
15. / Matheus Albuquerque full-stack developer draft illustration enthusiast open source lover web alchemist Felipe Borges back-ender minecraft lover mobile developer google enthusiast
16. / Matheus Albuquerque full-stack developer draft illustration enthusiast open source lover web alchemist Felipe Borges back-ender minecraft lover mobile developer google enthusiast
17. / Matheus Albuquerque full-stack developer draft illustration enthusiast open source lover web alchemist Felipe Borges back-ender minecraft lover mobile developer google enthusiast
18. / Matheus Albuquerque full-stack developer draft illustration enthusiast open source lover web alchemist Felipe Borges back-ender minecraft lover mobile developer google enthusiast
19. / Matheus Albuquerque full-stack developer draft illustration enthusiast open source lover web alchemist Felipe Borges back-ender minecraft lover mobile developer google enthusiast #after 
20. who ???
21. back-end development
22. back-end development devops
23. back-end development devops IaaS
24. back-end development devops IaaS PaaS
25. back-end development devops IaaS PaaS BaaS
26. back-end development devops IaaS PaaS BaaS FaaS
27. history
29. back in 2012…
33. problems we have
34. × development costs × scaling costs × operational management
35. so what ???
36. micro service nano service serverless Function as a Service Backend as a Service
37. #1 serverless
38. ✓ logic is executed in an environment without visible processes, OS, servers or virtual machines ✓ the responsibility for provisioning and managing the infrastructure belongs entirely to the service provider ✓ significantly depend on third-party services ✓ functionality needed was offered as services, covering databases, messaging, authentication, etc ✓ software developer focuses on writing code
39. #2 Backend as a Service
40. ✓ started as MBaaS ✓ provide server-side logic ✓ manage their internal states ✓ applications that do not have application- specific, server-side logic ✓ use third-party services for everything
41. ✓ Authentication as a Service (Auth0, AWS Cognito, Firebase) ✓ Logging as a Service (Loggly, Logsense, Amazon Elasticsearch Service) ✓ Analytics as a Service (Amazon Kinesis, Keen IO).
42. #3 Functions as a Service
43. ✓ short-living stateless functions ✓ triggered by events ✓ provider does the rest: ✓ provisioning instances of such functions, ✓ terminating instances, ✓ monitoring all of them, ✓ identity and logging services, ✓ …
44. Auth0 Webtask AWS Lambda Google Cloud Functions Azure Functions IronFunctions IBM OpenWhisk
45. #4 Micro & Nano services
46. ✓ a suite of small services ✓ each running in its own process ✓ each communicating with lightweight mechanisms ✓ these services are built around business capabilities ✓ independently deployable
47. microservice !→ all CRUD operations on an account nanoservice !→ each account operation: create, read, update, and delete
48. back to problems
49. #1 development costs
50. e.g. authentication
51. code your own authentication functionality
52. ✓ removes much of the database administration overhead ✓ provides mechanisms of authorization for different types of users
53. #2 scaling costs
54. 1 request / minute 50 ms / request 0.1% mean CPU usage / hour
55. 100ms compute / minute 0.15% of the overall time
56. ✓ democratize things ✓ breaking down components by logic / domain even without working with the operational costs (startups )
57. e.g. inconsistent traffic
58. https://martinfowler.com/articles/serverless.html
59. 200 https://martinfowler.com/articles/serverless.html
60. autoscaling
61. less than 4% of total uptime total hardware capability x 10
62. try: code optimization
63. #3 operational management
64. on the BaaS side
65. less components maybe… = less work *
66. on the FaaS side
67. × compressing your code × puppet / chef × start / stop shell scripts × decisions about whether to deploy one or many containers on a machine
68. when to use?
69. #1 slack bots
70. #2 database wrappers
71. #3 scientific computing
72. http://ericjonas.com/pywren.html
73. …
74. different ways
75. #1 command pattern
76. http://freecontent.manning.com/patterns-for-solving-problems-in-serverless-architectures/
77. #2 messaging pattern
78. http://freecontent.manning.com/patterns-for-solving-problems-in-serverless-architectures/
79. #3 priority queue pattern
80. http://freecontent.manning.com/patterns-for-solving-problems-in-serverless-architectures/
81. #4 fan-out pattern
82. http://freecontent.manning.com/patterns-for-solving-problems-in-serverless-architectures/
83. #5 pipes & filters pattern
84. http://freecontent.manning.com/patterns-for-solving-problems-in-serverless-architectures/
85. drawbacks
86. × not efficient for long-running applications × vendor lock-in × introduce additional overhead for function/microservice calls × multitenancy × cold start × different methods for logging in functions
87. up & running
89. $ npm install -g firebase-tools $ firebase init $ firebase deploy $ firebase login $ firebase init functions const functions = require('firebase-functions'); …
90. in the end…
91. ✓ reduced time ✓ lower operational and development costs ✓ works well with agile development ✓ allows developers to focus on code and to deliver faster ✓ fits with microservices ✓ reduces the complexity ✓ simplifies packaging and deployment
92. !!!
94. questions?
95. thanks! @ythecombinator @feliborgez